From Iran to Zaporizhzhia

War around nuclear facilities is no longer a remote contingency. It is an emerging test of whether international humanitarian law (IHL) can restrain military operations before nuclear safety systems are pushed beyond recoverable limits. Recent hostilities involving Iran and Israel, alongside the longer-running crisis around Zaporizhzhia in Ukraine, show that military activity is increasingly taking place in and around nuclear-related sites.

The danger is not confined to a direct strike on a reactor building. It also lies in the progressive degradation of the systems that keep nuclear risk contained. Zaporizhzhia illustrates this danger clearly. Repeated shelling, recurrent loss of off-site power, damage to radiation-monitoring systems, pressure on operating staff, military presence near reactor units, and the destruction of the Kakhovka dam’s wider water infrastructure have shown how nuclear danger can accumulate through disruption, intimidation, and infrastructural fragility, even without a catastrophic strike on the reactor vessel itself.

The Iranian context raises the same problem from a different angle. Even where the asserted military objective is counter-proliferation or strategic denial, attacks on or near nuclear-related infrastructure may create risks that cannot be confined to the target, the belligerents, or the duration of the operation. Once safety-critical systems are destabilised, the consequences may exceed the immediate battlespace, producing harms that are difficult to predict and harder to contain.

For that reason, the legal question is not only whether IHL prohibits a particular attack at the moment of targeting. It is also whether IHL can require forms of anticipatory restraint capable of preventing nuclear danger from becoming irreversible.

IHL is not silent on this problem. Article 56 of Additional Protocol I is the provision most directly aimed at preventing catastrophic civilian harm from attacks on works and installations containing dangerous forces. Yet its protection remains underinclusive. It singles out nuclear electrical generating stations for special protection, but leaves other nuclear-related facilities outside its core rule. It also contains an exception where a station provides electric power in “regular, significant and direct support” of military operations. In practice, that combination of narrow coverage and elastic exception can make the rule less stable than the risk warrants.

Existing IHL already requires serious attention to foreseeable indirect and cumulative civilian harm; that AP I Articles 56 and 57 support a plausible risk-differentiated interpretation in the nuclear context; and that states should make that interpretation operational by adopting hazard-based targeting guidance for military operations involving nuclear-related facilities.

The normative change proposed here is therefore practical rather than formally legislative. States should distinguish, in advance, between nuclear-related facilities according to the scale and foreseeability of radiological, environmental, and civilian harm that could follow from military action. Not all nuclear-related facilities create the same level of external danger.

Operating nuclear power plants, high-risk spent-fuel pools, and certain reprocessing or high-activity waste contexts belong at the top of the hazard spectrum and should be subject to a strong presumption of restraint grounded in precaution, proportionality, and the dangerous-forces logic of AP I. Lower-risk facilities remain governed by the ordinary rules which should be applied with much greater attention to indirect and cumulative effects on nearby higher-hazard sites. The argument is that IHL’s existing rules on dangerous forces, precautions, and proportionality should be translated into clearer operational thresholds, evidentiary requirements, and decision-making procedures where foreseeable nuclear risk is greatest.

Nuclear Risk Is Not Uniform

A central weakness in legal debate on this issue is the tendency to treat “nuclear facilities” as though they form a single class. They do not. Their risk profiles differ sharply, and any serious legal assessment must begin from that difference.

Operating nuclear power plants sit at the highest end of the hazard spectrum. They are not simply buildings or pieces of infrastructure. They are complex socio-technical systems that depend on reliable cooling, stable off-site power, functioning emergency systems, secure staffing conditions, and operational predictability. When those conditions are degraded, risk rises quickly. The danger is not limited to an obvious breach of containment. It can arise through the interaction of smaller disruptions that gradually push a facility towards a far more serious event as demonstrated in Zaporizhzhia. The danger there has never depended only on whether a reactor vessel is directly struck. It also turns on loss of off-site power, stress on emergency systems, personnel operating under coercive conditions, interruptions to maintenance, and military activity in the immediate environment of the plant. The risk is cumulative and systemic. It does not map neatly onto a narrow conception of “attack.”

The same point helps explain why nearby military activity matters even where no radiological release occurs. Reports of a projectile landing near Bushehr, if accurate, are legally significant because repeated operations around an operating nuclear power plant narrow the margin of safety. A legal approach that waits for visible damage before treating the situation as serious is already responding too late.

At the same time, it would be a mistake to treat every nuclear-related site as if it posed reactor-scale danger. Many research reactors operate at low power with smaller inventories. Their destruction can still be grave, but the likely consequences are often more localized than those associated with a large operating power reactor. Dry-cask storage also differs. An attack may cause serious local contamination and long-term remediation challenges, but wide off-site releases are generally less plausible absent extraordinary conditions.

Spent-fuel pools require more careful differentiation. In high-heat and high-density conditions, they can be extremely dangerous if water is lost or cannot be replenished. Yet not every pool presents the same hazard. Inventory, cooling time, design, and configuration matter. A functional approach is therefore more defensible than a blunt categorical one, and the same functional logic should apply to fuel-cycle facilities where high-activity materials or safety-critical dependencies create comparable risks.

This produces a simple operational premise: nuclear risk is unevenly distributed across facility types and, crucially, across safety functions. The legal question is how IHL should respond to that uneven distribution without pretending that ordinary targeting categories fully capture the danger.

What the Law Already Does – And Where It Falls Short

Existing law contains more than is sometimes acknowledged. In international armed conflict, Article 56 of AP I gives special protection to nuclear electrical generating stations where an attack may cause the release of dangerous forces and severe civilian losses. Additional Protocol II contains a parallel rule in Article 15 for non-international armed conflict. More generally, customary IHL demands particular care with works and installations containing dangerous forces. The ordinary rules of distinction, proportionality, and precautions also continue to apply. That means that even where Article 56 does not apply directly, attacks affecting nuclear-related facilities must still be assessed under the general targeting rules, including AP I Article 52 on military objectives, Article 51 on proportionality, and Article 57 on precautions.

Those general rules matter greatly in nuclear settings. Target verification, feasible precautions, and the choice of means and methods all require serious attention to foreseeable radiological and environmental consequences. They also require close attention to indirect effects.

Yet the existing legal architecture remains incomplete in three respects. First, Article 56 is limited in scope. Research reactors and much other nuclear-related infrastructure fall outside its special regime, even where attacks on those facilities may increase systemic danger at nearby higher-hazard sites. Secondly, the exception for “regular, significant and direct support” is too pliable if read loosely. Nuclear risk is too grave to permit broad strategic claims or general allegations of militarisation to satisfy that threshold. If the exception is interpreted generously, Article 56 risks becoming a language of justification rather than a real restraint.

Thirdly, the ordinary proportionality framework becomes strained in nuclear cases. Proportionality  presumes that expected civilian harm can be estimated with some reasonable degree of confidence. But in severe nuclear scenarios, uncertainty is often profound. Probability distributions, release pathways, long-term contamination, and intergenerational effects may not be knowable with precision in real time. That uncertainty matters legally because it affects both sides of the proportionality and precautionary analysis: it makes expected civilian harm harder to quantify, and it increases the importance of feasible alternatives, verification, and precautionary cancellation or suspension where the risk cannot responsibly be assessed.

Institutional limits deepen the problem. The nuclear safety and safeguards framework is built mainly for peacetime. The International Atomic Energy Agency (IAEA) can monitor, assess, warn, and assist. It cannot demilitarise a site or enforce protected zones during active hostilities. The IAEA’s seven pillars reinforce the systems-based nature of the problem: nuclear safety in armed conflict depends not only on physical integrity, but also on power supply, staffing, logistics, radiation monitoring, emergency preparedness, and reliable communications with regulators. Legality therefore depends heavily on how attacking states understand and internalise evidentiary burdens and feasible alternatives.

A Risk-Differentiated Reading of Existing IHL

A more operationally effective approach can begin with the law already in place. The claim is interpretive rather than legislative. IHL already requires distinction, proportionality, precautions, and special care around works and installations containing dangerous forces. The harder question is how those duties should operate where the foreseeable harm is not only immediate civilian injury, but the destabilisation of nuclear safety systems that may produce future, continuous, and geographically dispersed civilian harm. A serious radiological release, prolonged loss of cooling, or collapse of emergency-response capacity may generate consequences that continue after the attack, affect civilians far from the site, and engage legal regimes beyond IHL, including environmental protection, human rights, nuclear safety, and state responsibility.

Article 56’s exception should therefore be read narrowly and in an evidence-burdened way. General militarisation, occupation, or strategic importance should not suffice. A state invoking the exception should have to show, on a credible and facility-specific basis, that the particular safety-critical systems whose impairment would create catastrophic danger are being used to provide “regular, significant and direct” support to military operations. That does not eliminate the exception. It simply makes it workable and prevents it from collapsing into broad narrative claims advanced after the fact.

Article 57’s precautionary duties should also be made concrete in nuclear settings. A serious application of feasibility should create a safety-critical no-go logic around core cooling, spent-fuel cooling, containment integrity, emergency power, and off-site power interfaces at the highest-risk facilities. Those functions should be treated as subject to a strong presumption against attack or impairment unless the exception is established with credible evidence and no equally effective feasible alternative exists. It means that, where loss of a safety-critical function could produce catastrophic civilian consequences, feasibility must be assessed with particular rigour: through stricter verification, more serious consideration of alternatives, and greater attention to harm minimisation.

This is where reversibility matters. In many nuclear contexts, the key legal question is whether a military objective can be achieved without crossing irreversible safety thresholds. Where temporary, reversible, or non-kinetic measures can achieve the required military effect with materially lower hazard, feasibility should be read to favour them. Reversibility is part of what serious precaution requires when the downside risk is catastrophic.

The same logic also demands evidentiary discipline. In this field, the law cannot rely on post hoc reassurance that no radiological consequence occurred. What matters is whether the foreseeable risk was responsibly assessed before force was used. At a minimum, any operation implicating high-hazard nuclear contexts should rest on a documented pre-strike assessment that sets out the hazard assumptions, plausible release pathways, feasible alternatives, and evidentiary basis for invoking any exception. That does not create a new adjudicative body, it creates a standard against which legality can later be reviewed. More importantly, it changes incentives in advance by making superficial justifications less comfortable. Framed this way, documentation is not an additional treaty obligation; it is a practical implication of serious Article 57 compliance in a setting where legality depends on what was reasonably foreseeable before the attack.

Modest institutional bridges would make this restraint more administrable. The goal is not to transform the IAEA into a wartime enforcement authority. It is to support restraint through limited measures: notification practices, rapid technical assessments where feasible, and pre-agreed safety-zone templates. Existing practice shows that such restraint is not politically impossible. India and Pakistan already maintain a bilateral non-attack arrangement for nuclear installations, supported by an annual exchange of facility lists. The Pelindaba Treaty likewise prohibits parties from taking, assisting, or encouraging armed attacks against nuclear installations in the African nuclear-weapon-free zone. These examples are imperfect, but they show that states can accept practical restraint mechanisms when nuclear risk is understood as a shared vulnerability.

Conclusion

International law already recognises that nuclear danger is different in kind. But it has not yet aligned its legal tools closely enough with the actual gradient of risk across nuclear facility types and safety functions. That gap is a reason to apply IHL with greater precision.

A strong presumption of restraint for the highest-hazard settings, combined with graduated but still rigorous control elsewhere, would make the application of proportionality and precaution more realistic and disciplined. Read together, the dangerous-forces logic of Article 56 and the feasibility requirements of Article 57 support a more concrete rule of conduct: safety-critical systems at operating nuclear power plants and high-risk spent-fuel pools should not be targeted unless the exception is established on a credible and facility-specific evidentiary basis and no equally effective feasible alternative exists. Reversible options should be preferred where they can achieve the military aim with materially lower hazard. At a minimum, decisions should be grounded in documented pre-strike assessment rather than broad ex post justification.

That approach would narrow exception elasticity, strengthen evidentiary discipline, and better connect IHL’s rules to the way catastrophic nuclear harm actually becomes possible. As military operations increasingly touch nuclear-related sites, this is no longer a marginal issue. It is a test of whether IHL can still operate as a meaningful constraint at one of war’s most dangerous frontiers.